Communication Technology

 

 

 

Distributed MIMO

 

Distributed MIMO is a technique that utilizes distributed antennas across different radio devices to enhance communication capabilities. This approach allows for the utilization of a large number of antennas that are distributed rather than being placed in a single array, as seen in traditional collocated-MIMO configurations. Distributed MIMO has been shown to offer advantages in ultra-high-density user environments, such as stadiums and crowded urban areas, by providing large-capacity transmission and maintaining better data transmission compared to conventional collocated-MIMO setups .

The distinctive aspect of Distributed MIMO lies in the spatial distribution of these antennas across multiple, separate locations rather than being centralized at a single device or location. This distribution enables the network to exploit spatial diversity more effectively, enhancing signal quality, reliability, and data throughput rates.

Cooperative MIMO, also known as network MIMO or virtual MIMO, is another technology closely related to distributed MIMO. It involves grouping multiple devices into a virtual antenna array to achieve MIMO communications, thereby improving capacity, cell edge throughput, coverage, and group mobility of wireless networks in a cost-effective manner. While cooperative MIMO increases system complexity and signaling overhead, its benefits include enhanced system capacity through decorrelating MIMO subchannels and exploiting the advantages of distributed antennas.

Source : Distributed MIMO (ResearchGate)

  • Four antenna arrays surround the user, labelled AntArray with the subscripts i, j, k and l. Each one is drawn as a red panel carrying a grid of blue dots, and every dot is one antenna element.
  • The blue arrow leaving each panel is labelled perpend with the same subscript. It is that array's boresight, and the four of them point in different directions.
  • The orange dashed lines run from the person in the middle to all four arrays, so every array serves the same user at the same time.
  • The x, y and z axes drawn at the corner of each panel give that array its own coordinate frame.

One detail of the drawing matters more than it first appears. Each of the four is itself a collocated array rather than a single antenna, so distributed MIMO here means a set of small arrays spread out, and not a scattering of individual elements.

The separate coordinate frames are the reason network planning appears in the challenge list further down. A channel model has to translate the user position into each array's own frame before it can work out an angle of arrival for that array.

Why spreading the antennas helps

The two sections below list what distributed MIMO offers and what it costs. Neither list says where the gain comes from, and that answer is worth having first, because all three parts of it follow from geometry rather than from signal processing.

Distance is the first part. Path loss grows with distance, and a collocated array puts every one of its elements at the same distance from the user. Spreading the elements means the nearest of them is much closer than that average, and the nearest link dominates the received power.

The arithmetic is worth doing once. A user 200 m from a single site might be 50 m from the closest of four distributed sites. A path loss exponent near 3.5 turns that four-fold reduction in distance into roughly 21 dB. No processing gain on this page comes close to 21 dB.

Shadowing is the second part. A building between the user and the array removes the link, and every element of a collocated array sits behind the same building. Separated arrays are blocked by different objects, so the chance that all of them are blocked at once is far smaller.

That effect has a name, macro diversity, and better equipment at a single site never produces it. Adding antennas to a shadowed array leaves it shadowed.

Channel conditioning is the third part. MIMO capacity depends on the rank of the channel matrix rather than on the antenna count alone. Elements half a wavelength apart on one panel see almost the same channel when the scattering is poor. The matrix then loses rank and the extra streams carry nothing.

Arrays metres or hundreds of metres apart never meet that problem. Their channels differ because their positions differ, so the streams stay separable even on a clear line of sight path where a collocated array would not.

The figure above says the same thing in a different form. Four arrays facing four ways see one user along four distinct paths, and it is the distinctness rather than the count that produces the gain.

Advantages

Distributed MIMO is particularly useful in scenarios where high data rates and reliability are crucial, such as in cellular networks, wireless broadband services, and in applications requiring extensive coverage and capacity, like smart cities and large public venues. Its development and deployment are part of the broader evolution towards more efficient, reliable, and high-capacity wireless communication systems, including next-generation networks like 5G and beyond.

Here are the key aspects and benefits of Distributed MIMO:

  • Spatial Diversity: By using antennas spread out over a wide area, distributed MIMO systems can significantly reduce the likelihood of signal fading and improve the robustness of wireless communication. This spatial diversity helps in combatting the detrimental effects of multipath propagation where signals take multiple paths to reach the receiver, causing interference and signal degradation.
  • Increased Capacity: Distributed MIMO systems can increase the network capacity by allowing the simultaneous transmission and reception of multiple data streams. This is achieved through advanced signal processing techniques that exploit the spatial dimensions of the communication channel.
  • Improved Coverage: By strategically placing antennas across a geographic area, distributed MIMO can extend the coverage area of a wireless network. It enables efficient signal transmission to areas that might be difficult to reach with traditional centralized antenna systems.
  • Interference Management: These systems can also offer improved interference management. By using spatial filtering techniques, a distributed MIMO system can focus the transmission power towards intended users while minimizing interference to others.
  • Scalability: Distributed MIMO networks are inherently scalable, allowing for the addition of more antennas and nodes to improve performance and coverage as needed without a complete overhaul of the existing infrastructure.
  • Energy Efficiency: By enabling more direct communication paths and leveraging advanced signal processing, distributed MIMO systems can be more energy-efficient compared to traditional approaches, especially in large-scale deployments.

Challenges

While Distributed MIMO technology offers significant benefits in terms of increased capacity, coverage, and reliability, it also presents several challenges that need to be addressed for its effective deployment and operation.

Addressing these challenges requires ongoing research, development, and innovation in wireless communication technologies, as well as collaboration among industry, academia, and regulatory bodies. Advances in signal processing, network design, and spectrum management, along with the development of cost-effective hardware, are essential for realizing the full potential of Distributed MIMO in future wireless networks

Here are some of the key challenges:

  • Complexity in Signal Processing: The use of multiple, distributed antennas for transmitting and receiving signals introduces complexity in signal processing. Advanced algorithms are required to efficiently combine and decode signals from multiple paths, which can increase the computational load and latency.
  • Synchronization: Achieving precise synchronization among distributed antennas is critical for the effective operation of a Distributed MIMO system. Timing and phase synchronization must be maintained to ensure that signals are coherently combined, which can be challenging, especially in dynamic environments with moving users and varying channel conditions.
  • Channel Estimation and Feedback: Accurate channel estimation is crucial for optimizing the performance of MIMO systems. However, in a distributed setup, obtaining accurate channel state information (CSI) becomes more complex due to the increased number of transmission paths and the variability of the wireless channel. Furthermore, the overhead associated with feeding back this information from receivers to transmitters can be substantial.
  • Interference Management: While Distributed MIMO can help manage interference through spatial filtering, the distributed nature of the system can also introduce new interference scenarios, especially in densely deployed networks. Designing effective interference management strategies that adapt to changing network conditions is a significant challenge.
  • Network Planning and Deployment: Planning and deploying a Distributed MIMO system involves strategically positioning antennas to maximize coverage and performance while minimizing interference. This requires sophisticated network planning tools and methodologies, as well as considerations for the physical and regulatory environments in which the antennas are deployed.
  • Scalability and Flexibility: Ensuring that Distributed MIMO systems are scalable and flexible to adapt to varying network demands and configurations is challenging. The system needs to be designed to easily incorporate additional antennas and adapt to changes in network topology and usage patterns without significant reconfiguration.
  • Energy Consumption: Although Distributed MIMO has the potential to be more energy-efficient than traditional systems, managing the energy consumption of multiple distributed antennas and the associated signal processing can be challenging, especially in large-scale deployments.
  • Cost: The deployment of multiple antennas and the requisite infrastructure for Distributed MIMO can involve significant costs. Reducing the cost of hardware, installation, and maintenance is crucial for making Distributed MIMO economically viable for wide-scale adoption.
  • Regulatory and Spectrum Issues: Deploying Distributed MIMO systems may face regulatory challenges, including spectrum allocation and management, as well as adherence to emission standards and interference regulations.

Coherent and non-coherent operation

Two entries in the list above, synchronization and channel estimation, are one problem seen from two sides. Whether the separated radios can act as a single array depends on it, and the answer splits distributed MIMO into two modes that perform very differently.

Coherent joint transmission is the ambitious mode. Every array sends the same data with its phase chosen so the signals arrive at the user in step, and the amplitudes add before the squaring rather than the powers adding afterwards. N arrays then deliver N2 times the power of one.

Non-coherent joint transmission is the modest mode. The arrays transmit with no common phase reference, so the powers add and N arrays deliver N times the power. The gain is real, it is smaller, and it needs almost nothing from the network.

What coherence demands is a shared sense of phase. Every radio runs its own oscillator, and two free-running oscillators drift apart, so the phase relationship the transmitter assumed stops being true.

The tolerance is tighter than it sounds. A phase error of a quarter wavelength turns constructive addition into partial cancellation, and at 3.5 GHz a quarter wavelength is about 21 mm. A mismatch in cable length alone can produce it.

Three fixes are in use. A clock distributed over the fronthaul is the cleanest, a GNSS receiver at each site is the easiest to deploy, and synchronisation over the air between the radios themselves avoids both. The BeamSync paper in the reference list takes the third route.

Reciprocity calibration is the other half of the problem, and it belongs to TDD. The downlink channel is inferred from the uplink measurement, which only holds if the transmit and receive chains have matching responses, and they do not. Their amplifiers and filters differ.

A calibration factor per chain fixes that, and distributing the radios makes it harder because the chains no longer share a board. The ResearchGate paper in the reference list is about that exact problem, and its title names the method.

The practical result is a per user decision. A slow moving user served by a well synchronised group gets coherent transmission, and a fast moving one, or one served across a group that cannot hold phase, gets the non-coherent fallback.

What the fronthaul has to carry

Cost and deployment both appear in the challenge list, and one component dominates both of them. The antennas are cheap. The link that carries their signals back to wherever the joint processing happens is not.

Joint processing means the samples from every array have to meet somewhere. A central unit computes the combining weights, so either the samples travel to it or part of the processing moves out to the radios.

Sending raw samples is expensive, and the number is easy to compute. One 100 MHz carrier sampled at 122.88 Msample/s, with I and Q at 15 bits each, is about 3.7 Gbit/s for a single antenna port. Four ports on one array is roughly 15 Gbit/s before any overhead is added.

Worse, that rate never falls. A raw sample stream costs the same whether the cell is carrying traffic or sitting idle, because it describes the waveform rather than the data.

Moving part of the physical layer into the radio changes that. The fronthaul then carries frequency domain data for the resource blocks actually in use, and traffic scales with load rather than with bandwidth. O-RAN's split 7-2x is the usual choice, and the RAN Architecture page covers where the splits fall.

Delay is the second constraint and it is the one that sets the size of the system. The combining has to finish inside the HARQ timeline, so the fronthaul gets a budget of tens of microseconds rather than milliseconds.

Fibre carries light at about 5 microseconds per kilometre, so that budget puts the central unit within roughly ten to twenty kilometres of the radios. Arrays further apart than that cannot join the same coherent group, whatever the link capacity.

Those two limits together are why distributed MIMO is deployed as clusters rather than as one network-wide array. The cluster is as large as the delay budget and the fibre plan allow, and no larger.

Cell free massive MIMO

Research literature mostly calls this architecture something else, and anyone reading further will meet the other name immediately. Cell free massive MIMO is distributed MIMO taken to the point where cells stop existing.

The setup is simple to state. Many access points, each carrying a few antennas, all connect to one central unit, and every access point serves every user. No access point owns a cell, so no user stands at a cell edge.

Removing the cell edge is the whole argument. The worst place to stand in a cellular network is midway between two sites, where the wanted signal is weak and the interference is strong. If both sites serve you, the interferer becomes a second transmitter instead.

Scalability is the catch, and it appears in two places. Serving every user from every access point costs twice over. The central unit computes weights across a matrix that grows with the whole network, and the fronthaul carries everything everywhere.

User centric clustering is the standard fix. Each user is served by the subset of access points that reach it well, usually a handful of them, and the subsets overlap rather than tile. The system keeps the cell free behaviour without the cell free cost.

How much processing to centralise is the remaining choice. Doing all of it at the central unit gives the best performance and the heaviest fronthaul. Processing locally at each access point, and combining only the soft estimates centrally, gives most of the gain for a fraction of the traffic.

That is the same trade the section above describes, arriving from the other direction. Where the processing sits decides what the fronthaul has to carry, and what the fronthaul can carry decides where the processing may sit.

Status is worth stating plainly. Cell free massive MIMO is a laboratory and trial technology rather than a deployed one, and the NEC and Nokia entries in the reference list are demonstrations of that kind rather than products.

Where this appears in 3GPP

The specifications carry no feature called distributed MIMO, and a reader searching them for that name finds nothing. What they carry instead is a run of features that each do part of the same job, under other names.

CoMP came first, in LTE Release 11. Coordinated multipoint defined joint transmission, dynamic point selection, and coordinated scheduling with beamforming, and its joint transmission mode is coherent distributed MIMO under another name.

CoMP was rarely deployed, and the reason is the section above. The backhaul of 2012 could not carry the samples or meet the delay budget, so the feature existed in the specification and not in the field.

Multi-TRP is the current form, from NR Release 16 and 17. A UE can be configured to receive from more than one transmission and reception point, and the Enhanced Massive MIMO page covers how that is set up.

Two flavours exist and the difference matters. Under single-DCI one scheduling message covers both points and the UE treats them as one transmission, while under multi-DCI each point schedules on its own. The single-DCI form is the closer relative of joint transmission.

The TCI state is the mechanism underneath. A TCI state ties a transmission to a reference signal the UE has already measured, so configuring two of them is how the network tells the UE that two spatial sources are being used.

One deployment needs no feature at all. Radio units spread through a building and fed from one baseband unit look to the UE like a single cell, which is the oldest form of this idea and still the most widely installed.

What the specifications do not describe is the network-wide coherent array. Nothing in them defines it, the fronthaul that would carry it lives in O-RAN rather than in 3GPP, and the research name for it is the one the section above gives.

YouTube

Refernces